OpenAI announced a limited preview of the GPT-5.6 series, introducing Sol, Terra, and Luna as its newest family of AI models.
GPT-5.6 Sol is OpenAI’s flagship model in the series. Terra is positioned as a balanced model for everyday work, while Luna is designed as a faster and more affordable option.
OpenAI said Terra delivers competitive performance compared with GPT-5.5 at half the cost, while Luna brings strong capabilities at the company’s lowest cost.
The preview is initially being made available to a small group of trusted partners before a broader release. OpenAI said it plans to make GPT-5.6 Sol, Terra, and Luna generally available in the coming weeks.
The company said GPT-5.6 Sol launches with its most robust safety stack to date. OpenAI strengthened protections for higher-risk activity, sensitive cyber requests, and repeated misuse, while spending multiple weeks testing weaknesses and hardening the system against real-world attacks.
OpenAI said GPT-5.6 Sol is its strongest model yet and highlighted improved agentic capabilities across coding, biology, and cybersecurity. The company also introduced a new “max” reasoning effort that gives Sol more time to reason deeply.
The GPT-5.6 series also introduces an “ultra” mode that goes beyond the capabilities of a single agent by using subagents to accelerate complex work.
For coding workflows, OpenAI said GPT-5.6 Sol sets a new state of the art on Terminal-Bench 2.1, a benchmark that evaluates command-line workflows requiring planning, iteration, and tool coordination.
The company said GPT-5.6 Sol also shows improvements in biology workflows. On GeneBench v1, which evaluates long-horizon genomics and quantitative biology analyses, Sol achieved stronger results than GPT-5.5 while using fewer tokens.
OpenAI also described GPT-5.6 Sol as its most capable model yet for cybersecurity. The company said the model shifts the performance-efficiency frontier for long-horizon security tasks, including vulnerability research and exploitation, while maintaining stronger safeguards.
OpenAI said GPT-5.6 Sol is better at helping people find and fix vulnerabilities than reliably carrying out end-to-end attacks. The company also said the model does not cross the Cyber Critical threshold under its Preparedness Framework.
The GPT-5.6 preview uses layered safeguards, including model-level refusals for prohibited cyber assistance, real-time checks during generation, account-level review signals, differentiated access, monitoring, enforcement, and ongoing testing.
During the preview, GPT-5.6 models will initially be available through the API and Codex to select trusted partners and organizations. OpenAI plans to make the models more broadly available to users of ChatGPT, Codex, and the API soon.
OpenAI also introduced a new naming system with GPT-5.6. The number identifies the model generation, while Sol, Terra, and Luna identify capability tiers that can evolve on their own cadence.
GPT-5.6 pricing is set across the three model sizes on a per-1-million-token basis. Sol is priced at $5 for input and $30 for output; Terra at $2.50 for input and $15 for output; and Luna at $1 for input and $6 for output.
The company is also launching GPT-5.6 Sol on Cerebras in July at up to 750 tokens per second, with initial access limited to select customers as capacity expands.
